|
Title |
Experimental Validation of a Deep Learning Distortion-Corrected Laser-Vision System for Bridge Deflection Measurement
|
|
Authors |
이규완(Kyu-Wan Lee) ; 김도균(Do-Kyun Kim) ; 박영식(Young-Sik Park) |
|
DOI |
https://doi.org/10.11112/jksmi.2025.29.4.10 |
|
Keywords |
교량 변위; 딥러닝; 왜곡 보정; 비전; 레이저빔 Bridge deflection; Deep learning; Distortion correction; Vision; Laser beam |
|
Abstract |
This study proposes an improved non-contact sensor system for measuring structural displacement. While traditional contact-type sensors such as LVDTs offer high accuracy, they are difficult to install and may suffer from wear or deformation over long-term use. On the other hand, non-contact sensors like LDS can perform measurements without physical contact but often lack sufficient accuracy. To address these limitations, this research introduces a non-contact displacement measurement method utilizing a laser beam and a webcam. An economical multi-point bridge deflection measurement system was developed based on this approach. Furthermore, an AI-based high-precision measurement technique was implemented to enhance both the accuracy and efficiency of displacement estimation. The trained model achieved a maximum and minimum prediction error of 0.0484 mm and 0.0854 mm for the X-coordinate, and 0.0625 mm and 0.0596 mm for the Y-coordinate, respectively.
|